Clinical characteristics of bacterial infections in patients with decompensated cirrhosis and construction and verification of a risk prediction model

Background Bacterial infections and disease progression are correlated in individuals with decompensated cirrhosis. We aimed to construct and verify a risk prediction model for bacterial infections in patients with decompensated cirrhosis. Methods Retrospectively, 588 patients with decompensated cirrhosis treated at the First Affiliated Hospital of Hebei North University were selected as the training set, and 224 patients with decompensated cirrhosis treated at Zhangjiakou Traditional Chinese Medicine Hospital were selected as the validation set. The participants were divided into infected and non-infected groups according to whether they had bacterial infection or not. Clinical data were collected before the positive culture results; the variables were screened by least absolute shrinkage and selection operator (LASSO) regression. Multivariate regression was used to analyze infection risk factors to construct a nomogram model; the predictive effect of the model was evaluated by the area under the receiver operating characteristic curve (AUC). Calibration and decision curves were used to evaluate the model’s clinical application value. Results Bacterial infections occurred in 34.9% of patients; peritonitis was the main infection. Escherichia was the most cultured infectious agent among 68 pathogenic bacterial strains. Multivariate logistic regression analysis showed that independent risk factors for bacterial infections (P < 0.05) were electrolyte and acid-base imbalance, renal function impairment, liver failure, abnormal platelet (PLT) counts, a high Child–Pugh score, and a high Model for End-Stage Liver Disease (MELD) score. The AUCs of the predicted model were 0.802 in the training cohort and 0.832 in the validation cohort. Hosmer–Lemeshow tests showed a good fit between the model and verification groups. Decision curve analysis and calibration curves showed a high value for the prediction model. Conclusions The nomogram model showed favorable differentiation and prediction of bacterial infection risk and might be able to identify high-risk patients early.